Human behavior recognition method based on second order motion description operator

Human behavior trajectory is often submerged in complex background, and it is easy to be disturbed. The traditional dense trajectory behavior recognition method has high track redundancy, so it has a large amount of calculation. A human behavior recognition method based on foreground trajectory and motion differential descriptor is proposed. Firstly, optical flow method is used to estimate the amplitude of moving object in each frame, and then the foreground region is found. Only the foreground region related to behavior is extracted. In order to better describe the relative time information between different behaviors, a motion difference descriptor is introduced to describe the foreground trajectory. By calculating the direction information of the unit time motion difference of the trajectory points, the motion difference direction histogram is constructed. Then, the histogram features are encoded by Fv (Fisher vector) to form feature vectors. Finally, the behavior is classified by support vector machine (SVM). The experimental results on the benchmark data set show that the proposed method can extract human motion trajectory better, and can improve the recognition accuracy compared with the traditional methods.